For　the　purpose　of　handling　noisy　data　set　more　effectively,　an　approach　of　improvement　of　FOIL　system　based　on　GDT　is　proposed.　This　approach　introduces　the　GDT　and　rule　strength　technique　to　improve　the　performance　of　FOIL　system　at　the　aspect　of　learning　noisy　data　sets.　Through　explaining　an　example　using　the　improvement　algorithm,　we　analyze　the　result　which　shows　that　the　approach　is　an　effective　noise-tolerant　first-order　learning　system.